Evaluation of bioaerosol samplers for the detection and quantification of influenza virus from artificial aerosols and influenza virus–infected ferrets

Abstract Background Bioaerosol sampling devices are necessary for the characterization of infectious bioaerosols emitted by naturally‐infected hosts with acute respiratory virus infections. Assessment of these devices under multiple experimental conditions will provide insight for device use. Objectives The primary objective of this study was to assess and compare bioaerosol sampling devices using a) an in vitro, environmentally‐controlled artificial bioaerosol system at a range of different RH conditions and b) an in vivo bioaerosol system of influenza virus‐infected ferrets under controlled environmental conditions. Secondarily, we also sought to examine the impact of NSAIDs on bioaerosol emission in influenza virus‐infected ferrets to address its potential as a determinant of bioaerosol emission. Methods We examined the performance of low and moderate volume bioaerosol samplers for the collection of viral RNA and infectious influenza virus in vitroand in vivo using artificial bioaerosols and the ferret model of influenza virus infection. The following samplers were tested: the polytetrafluoroethylene filter (PTFE filter), the 2‐stage National Institute of Occupational Safety and Health cyclone sampler (NIOSH cyclone sampler) and the 6‐stage viable Andersen impactor (Andersen impactor). Results The PTFE filter and NIOSH cyclone sampler collected similar amounts of viral RNA and infectious virus from artificially‐generated aerosols under a range of relative humidities (RH). Using the ferret model, the PTFE filter, NIOSH cyclone sampler and the Andersen impactor collected up to 3.66 log10copies of RNA/L air, 3.84 log10copies of RNA/L air and 6.09 log10copies of RNA/L air respectively at peak recovery. Infectious virus was recovered from the PTFE filter and NIOSH cyclone samplers on the peak day of viral RNA recovery. Conclusion The PTFE filter and NIOSH cyclone sampler are useful for influenza virus RNA and infectious virus collection and may be considered for clinical and environmental settings.


| INTRODUC TI ON
Influenza virus remains a public health concern due to associated seasonal burden of disease and pandemic potential. 1,2 Personto-person transmission occurs by direct and indirect contact, by droplets (particles ≥10 μm) and potentially through droplet nuclei (particles ≤5 μm) which may be inhaled into the lower airways. [3][4][5][6] Lower respiratory tract infection is associated with increased disease severity and mortality compared to infection of the upper respiratory tract. 7,8 Understanding determinants of influenza A virus (IAV) bioaerosol-mediated transmission including environmental factors such as RH and temperature is relevant to mitigating spread. 9,10 RH can influence virion size, infectivity, and bioaerosol sampler performance. Early work with IAV determined that infectivity was best maintained at lower RH conditions. 11 Noti et al also conducted experiments using simulated coughs and demonstrated that IAV aerosol infectivity was maximized at low RH. 12 Furthermore, Lowen et al suggested that IAV transmission is highest at low RH, moderate at high RH (65%), and lowest at an intermediate RH (50%) using the guinea pig model. 10 In addition, the range of RH may vary significantly depending on climate (temperate vs tropical) and setting (health care vs agricultural or wet markets). The overall effect of RH on IAV infectivity and transmission is still under investigation, but the effective performance of bioaerosol sampling devices at a range of RH is important for characterizing virus-laden bioaerosols under different conditions. An area of IAV research currently lacking attention is the effect of widespread non-steroidal anti-inflammatory drugs (NSAIDs) and other antipyretic use on IAV emission by infected hosts. Early experiments indicated that influenza virus-infected ferrets treated with antipyretic compounds experienced increased nasal shedding, potentially leading to increased risk of transmission. 13  Assessing the burden of influenza virus in the air is technically and operationally challenging in real-world settings such as healthcare institutions. [15][16][17] Clinical and environmental studies have TA B L E 1 Summary of characteristics, inclusion rationale, and recommendations for the polytetrafluoroethylene (PTFE) filter, NIOSH cyclone sampler, and Andersen impactor  21,32 utilized a range of instruments to determine the risk of exposure to virus-laden bioaerosols in health care and agriculture. [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32] Limited study sizes underscore the need for consistent approaches across studies in similar settings in order to generate robust comparative data to form clearer conclusions. Bioaerosol sampling devices are essential to the investigation and characterization of IAV bioaerosol emissions and transmission. Many sampling devices have been employed to collect a number of different pathogens but were not explicitly designed for the collection of viruses and preservation of viral infectivity. The selection of a suitable bioaerosol sampling device is challenging since collection efficiency is influenced by the pathogen in question, sampler flow rate, the environment, sampling time, and other technical and operational aspects of each device.
Other factors such as cost and ease of use also influence device selection. Currently, there is no prevailing standard for the selection of a bioaerosol sampling device.
Filter-based bioaerosol sampling devices, cyclone samplers, and cascade impactors are the most commonly used and accessible instruments for recovery and detection of viruses, though none were explicitly designed for this purpose. 17,21,22,33 Filter-based instruments use membranes to collect particles as air is drawn through, whereas cyclone samplers recover particles by inertia, and some are capable of size fractionation. 15,33 Cascade impactors collect particles on solid or liquid media and may provide size fractionation as well. 15,34 These three portable sampler types have potential for implementation in standard practice for assessing the burden of virus in the air.
The primary objective of this study was to assess and compare bioaerosol sampling devices (Table 1)  conditions. Secondarily, we also sought to examine the impact of NSAIDs on bioaerosol emission in influenza virus-infected ferrets to address its potential as a determinant of bioaerosol emission.

| Ethics statement
Animal experiments were completed in a biosafety level 2+ containment facility at the Sunnybrook Research Institute (SRI, Toronto, Canada) in compliance with guidelines set by the Canadian Council on Animal Care and with approval of the SRI animal care committee.  system. RH and temperature were monitored throughout the experiments. RH within the chamber was maintained using a portable humidifier (Boneco), and temperature was maintained within a range of 19-23°C. Three different RHs were tested, low (<25%), medium (47%-53%), and high (78%-83%; n = 3 for each). Sampling devices were placed inside the chamber, except for the PTFE filter at high RH due to filter saturation. Thus, sampling with the PTFE filter during high RH conditions was conducted from outside the chamber via connecting tubing. The chamber was cleaned with ethanol and purged with water for 25 minutes before each nebulization/sampling event.

| Sample processing and analysis
Bovine serum albumin (BSA, 0.5%) was added to nasal wash sample supernatants after centrifugation at 800 g for 5 minutes at 4°C.
The PTFE filter was vortexed with 2 ml VTM for 1 minute, and the NIOSH cyclone sample tubes and filter were vortexed with 2, 0.5, or 1 mL of VTM for 1 minute for stages 1, 2, and the NIOSH filter, respectively. All samples were stored at −80°C after processing. Viral

| Statistical analysis
All data were initially tested for normality using the D'Agostino-Pearson omnibus normality test. Normally distributed data were tested using either an unpaired t test or one-way ANOVA to determine statistical significance. Following ANOVA testing, Tukey's multiple comparisons test was utilized to determine groups that were statistically different. Non-normally distributed data were assessed using either the Mann-Whitney test or the Kruskal-Wallis test to determine statistical significance (GraphPad Prism 6 software). The ratio of viral RNA to infectious virus was determined by dividing the viral RNA collected (copies) by infectious virus collected (PFU).

| PTFE filter and NIOSH cyclone samplers collected IAV RNA and infectious virus from artificial aerosols under a range of RH
To evaluate the ability of the PTFE filter and NIOSH cyclone sampler to collect influenza virus RNA and infectious virus, we aerosolized H1N1 and H3N2 influenza viruses into a custom-built artificial aerosolization chamber ( Figure 1A). The Andersen impactor was not used in this setting because the flow rate was too high relative to the low volume of the chamber. The PTFE filter and NIOSH cyclone samplers collected similar quantities of H1N1 and H3N2 RNA and infectious virus under all RH conditions ( Table 2). The ratio of viral RNA to infectious virus was determined to assess the loss of infectivity during the aerosolization, transit, and collection 17 and this was similar to both H1N1 and H3N2 influenza viruses (Table 2).
RH may affect viral viability in aerosols and bioaerosol sampler collection; thus, we also sought to compare instrument performance under a range of RH conditions. The RH within the artificial aerosolization chamber was maintained throughout aerosolizations at low (<25%), medium (47%-53%), or high (78%-83%) RH conditions. There was no significant difference between bioaerosol sampler collection at different RH conditions for RNA or infectious virus (P > .05, Table 1) except significantly less H3N2 RNA was collected by the NIOSH cyclone sampler at medium RH compared to low RH (P = .032; Table 2). Figure 2 indicates the size distribution of particles collected by the NIOSH cyclone sampler to provide further information on the influence of RH on influenza virus particle size. Similar particle sizes for both H1N1 and H3N2 influenza viruses were recovered under low and medium RH conditions using the NIOSH cyclone sampler, and the majority of infectious virus was recovered from the NIOSH filter (<1.0 μm) under these conditions (Figure 2). There were variable quantities of RNA and infectious virus from particles collected under high RH conditions.

| Collection of bioaerosols emitted by IAVinfected ferrets
Next, we sought to collect IAV using bioaerosol samplers from influenza virus-infected ferrets. We also sought to determine whether NSAID administration affected influenza virus-laden bioaerosol production in a mammalian model. Weight loss was similar between untreated and NSAID (meloxicam)-treated influenza virus-inoculated ferrets ( Figure 3A). Treated ferrets had significantly lower temperatures on day 1 and 7 p.i. (Figure 3B, P = .016 and P = .041, respectively), but no fevers were noted. Nasal wash viral loads were highest on day 1 p.i., reaching 6.20 ± 0.22 log 10 PFU/mL for untreated ferrets and 6.85 ± 0.26 log 10 PFU/mL for treated ferrets, and declined until termination ( Figure 3C) with no significant difference between untreated and treated animals (P > .05; Figure 3C).
The PTFE filter and NIOSH cyclone sampler collected the most viral RNA on day 3 p.i. for both untreated and treated ferrets ( Figure 4A, 4B). PTFE filters collected 2.26 log 10 copies/L air for untreated and 3.66 log 10 copies/L air for treated animals, while the NIOSH cyclone sampler collected 2.56 log 10 copies/L air for untreated and 3.84 log 10 copies/L air for treated ferrets on day 3 p.i. (Figure 4A, 4B). More viral RNA was collected from the air of treated ferrets than untreated ferrets, though statistical significance could not be determined due to small sample size ( Figure 4A, 4B).
Infectious virus was collected on day 3 p.i. from the PTFE filter sampling treated ferrets and from the NIOSH cyclone sampler sampling both untreated and treated ferrets ( Figure 4C). log 10 copies/L air were collected from untreated ferrets, and 6.09 log 10 copies/L air were collected from treated ferrets ( Figure 5A, 5B).
More viral RNA was collected by the Andersen impactor from ferrets treated with meloxicam than untreated ferrets, but statistical significance could not be determined ( Figure 5A, 5B) Low is <25% relative humidity, medium is 47%-53% relative humidity, and high is 78%-83% relative humidity.
b Viral RNA was determined by RT-qPCR and presented as copies of RNA per liter air sampled.
c SEM = standard error of the mean (n = 3 per relative humidity condition).
d Infectious virus was determined by plaque assay and presented as PFU per liter air sampled.
e Viral RNA to infectious virus ratio was determined by dividing viral RNA copies per liter air sampled by PFU per liter air sampled. *Significantly less H3N2 RNA recovered compared to the NIOSH cyclone sampler at low relative humidity (P < .05).
We also inoculated ferrets from untreated and meloxicamtreated groups and individually sampled each animal by nasal washing and using the PTFE filter ( Figure S1). Approximately 10.4% of aerosol samples were positive for IAV RNA and only 3 of 48 aerosol samples were positive for infectious IAV ( Figure S2). We attempted to determine a relationship between viral load and virus (RNA and infectious virus) collected from the air using a Fisher's exact test, chi-squared test, and a Spearman correlation but did not find a significant difference.  45,46 and to assess the environmental burden of avian influenza viruses in wet markets. 47,48 In this study, we evaluated three portable samplers using artificial aerosols and a translational, in vivo model.   counterparts. This indicates a potential enhancement of influenza virus bioaerosol production after NSAID treatment through an unknown mechanism, and could play a significant role during influenza epidemics and pandemics due to the widespread use of NSAIDs such as ibuprofen. 59 Transmission was not tested in this study since this

| CON CLUS ION
In summary, the Andersen impactor has a high flow rate, the ability to size fractionate particles, and is commercially available. The cumbersome nature and cost of the Andersen impactor impacts the feasibility for widespread clinical and field use, but this instrument may be ideal for collection of IAV RNA in experimental and other well-controlled settings. The NIOSH cyclone sampler is portable, lightweight, is able to size fractionate bioaerosols, collects IAV RNA and infectious virus, and works well under multiple RH conditions. It is not commercially available and requires time for processing and decontamination but can be used for personal sampling and whenever particle size frac-